Give your AI Agent control of your Chrome browser

Give your AI Agent (Claude Code, Kiro etc.) control of your Chrome browser using Chrome Dev Tools MCP server

• 3 min read •
#ai #claude-code #kiro

The Google Chrome team launched Chrome DevTools MCP server in September 2025. Since then, it has become easy to given an AI Agent (Kiro, Claude Code, Cursor, etc.) to the ability to use Chrome (to navigate pages, debug issues, take screenshots, take screen recordings etc.).

If you’d like to your AI agent the ability to use Chrome, follow the steps below.

First, as the chrome-devtools-mcp CLI’s README instructs, add Chrome DevTools MCP server to your AI Agent.

If you use Claude Code, add Chrome DevTools MCP server to the mcpServers field in Claude Code’s config file~/.claude.json

{
  "mcpServers": {
    "chrome-devtools": {
      "command": "npx",
      "args": ["-y", "chrome-devtools-mcp@latest", "--autoConnect"]
    }
  }
}

Alternatively, just run

claude mcp add chrome-devtools --scope user npx chrome-devtools-mcp@latest -- --autoConnect

If you use a version manager (mise, nvm etc.) to install node, switch to your default node version (say, node 24), install chrome-devtools-mcp

mise use node@24
npm install chrome-devtools-mcp

and then modify Claude Code to use the correct node version when starting the Chrome DevTools MCP server

{
  "mcpServers": {
    "chrome-devtools": {
      "command": "/Users/chunisan/.local/share/mise/installs/node/24/bin/node",
      "args": [
        "/Users/chunisan/.local/share/mise/installs/node/24/lib/node_modules/chrome-devtools-mcp/build/src/bin/chrome-devtools-mcp.js",
        "--autoConnect"
      ],
      "env": {
        "PATH": "/Users/chunisan/.local/share/mise/installs/node/24/bin:/usr/local/bin:/usr/bin:/bin"
      }
    }
  }

If you use Kiro, add Chrome Dev Tools MCP server to the mcpServers field in your Kiro agent's config file in ~/.kiro/agents/ If your Kiro agent has the useLegacyMcpJson property set to true, add Chrome Dev Tools MCP server to the mcpServers field in ~/.kiro/settings/mcp.json

Start Chrome, visit chrome://inspect/#remote-debugging and enable “Allow remote debugging for this browser instance”. If you’d like, you can open a different Chrome profile dedicated to your AI agent(s) before visiting chrome://inspect/#remote-debugging and enabling remote debugging.

That’s it! Open your AI Agent and ask it open any website in Chrome.

As AI agents are increasingly autonomous, having to allow remote debugging each time an AI agent uses Chrome is cumbersome and unproductive. To avoid this, start Chrome with a different user data directory and remote debugging enabled

/Applications/Google\ Chrome.app/Contents/MacOS/Google\ Chrome --user-data-dir="$HOME/Library/Application\ Support/Google/AI\ Google/ Chrome" --remote-debugging-port=9222 

If you’d like to have a different MacOS Dock icon for the above Chrome instance, install another Chrome instance from a different Chrome channel, Chrome Beta, Chrome Dev, Chrome Canary or Chrome for Testing or use the Mitosis app to install another Chrome instance

/Applications/Google\ Chrome\ for\ Testing.app/Contents/MacOS/Google\ Chrome --user-data-dir="$HOME/Library/Application\ Support/Google/AI\ Google/ Chrome" --remote-debugging-port=9222 

If time permits, I highly recommend reading chrome-devtools-mcp CLI’s README in entirety as it has many handy features (screen recording etc.).